JPS63182699A - Word reading information storage dictionary - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a dictionary that is used in a speech synthesis device and the like and stores word pronunciation information.

(Prior art) Conventionally, in text analysis processing for rule-based synthesis of speech, a word dictionary that stores information such as the orthographic notation of each word, 9 parts of speech, pronunciation, accent, etc. is used to analyze input text. . For example, Information Processing Society of Japan Japanese Document Processing Study Group Material 2-4 (1985) by Fukushima et al.
, ``Sentence Analysis for Sentence to Speech Conversion in Readers for the Blind'' uses kanji as headings, uses a dictionary with information such as pronunciations, accents, and parts of speech, and analyzes input text while referring to connection tables of parts of speech. It converts it into a string of words and outputs speech synthesis information consisting of information such as pronunciation and accent, and syntax information.

(Problems to be Solved by the Invention) However, in order to generate clear synthesized speech by regular speech synthesis, each phoneme, especially consonants, must be synthesized strongly and clearly. The reason why consonants are especially important is that vowels have long durations and high power, and there are few types of them, so they are less likely to be confused with each other, whereas consonants have low power and many types.
This is because they are easily confused with each other. However, in the conventional method of speech synthesis, when each phoneme, especially a consonant, is emphasized by increasing its strength or power or lengthening its duration in order to improve clarity, the - sound The problem was that because each note became stronger, it lacked smoothness, resulting in a harsh and unnatural synthesized sound, and it became tiring to listen to for a long time.

An object of the present invention is to provide a dictionary that solves this problem and provides information for instructing to emphasize consonants only when they are likely to be confused with each other.

(Means for Solving the Problems) Means provided by the present invention in order to solve the above-mentioned problems is a word reading information storage dictionary consisting of dictionary entries having phonetic transcriptions consisting of at least a series of phonetic symbols. , the phonetic transcription of a certain dictionary entry has the same phonetic symbol in many corresponding positions in the phonetic symbol sequence, and differs in some corresponding positions, compared to the phonetic transcription of some other dictionary item. In the case of having similar phonetic symbols, the certain dictionary entry is characterized in that information about the position of the different but similar phonetic symbol in the phonetic notation is added.

(Function) In the present invention, as a result of analyzing an input character string while referring to a dictionary, the text analysis section sends information such as pronunciation and accent of the word to the speech synthesis section. It sends information on whether there are any words or phrases that are easily confused with each other, and if so, which phonemes are likely to be confused with each other.Based on that information, the speech synthesis section adjusts the degree of emphasis on consonants, etc. The present invention provides a dictionary that makes it possible to synthesize speech under control.

For example, the word "hospitality" / ka is added to the dictionary.
Consider the case where N ta i/ and ``Cheers J /ka N pa i/'' exist.The expression surrounded by diagonal lines is the pronunciation of each word represented by a string of phonetic symbols.The consonant h/
Given that and /p/ are predefined as similar phonetic symbols, these two words each have similar phonetic symbols in the consonant part of the third syllable, and identical phonetic symbols in other positions. Therefore, if the marker indicating the existence of similar phonetic symbols is "*゛', the phonetic notation of these words will be expressed as /1caNt*ai/ and /ka N p*a i/, respectively.By synthesized speech, If the input character string to be output contains any of these words, for example "hospitality", the speech synthesis unit will
It becomes possible to selectively emphasize only the consonant /L/.

(Example) Hereinafter, the present invention will be described in more detail with reference to the drawings. FIG. 1 is used to create one embodiment of the present invention. FIG. 3 is a diagram showing input dictionary entries for "a word 1 hospitality" and "cheers"; 2 to 4 are flowcharts showing the flow of processing for adding information indicating the position of a similar phonetic notation of a dictionary item with similar pronunciation to the input dictionary item.

First, define a set of similar phonetic symbols (201).
For example, /l/ and /p/ and /, also /m/ and /n
/, and /b/, /d/, /g/, and /1-/ are similar phonetic symbols, then the set S of all similar phonetic symbols is s= ((/P / /l/ /ni/) (/m/
/n/) (/b/ /d/ /g/ /r/))
becomes.

Next, all input dictionary items are sorted by phonetic notation (202). This sorting may be performed, for example, in alphabetical order of phonetic symbols forming the phonetic notation.

However, in this case, similar phonetic symbols are considered to be the same and sorted. Therefore, for example, "hospitality" 101 and "cheers" 102 in FIG. 1 appear in a series of dictionary items with similar pronunciations. Here, dictionary entries with similar pronunciations are those that have phonetic symbols that are similar at some corresponding positions in the phonetic symbol sequence that constitutes the phonetic notation, and that have the same pronunciation at other corresponding positions in the phonetic symbol sequence. It refers to multiple dictionary entries that have symbols.

Through the processing up to step 202, the input dictionary items are rearranged so that dictionary items with similar pronunciations appear consecutively. Therefore, for each sequence of dictionary items with similar pronunciation,
The processing shown in FIGS. 3 and 4 is performed for each group as one group (204).

FIG. 3 shows the flow of processing for one group. By the definition of phonetically similar dictionary entries, each dictionary entry within a group has the same number of phonetic symbols in its phonetic transcription. Let this number be MAX-PHONE (301
). The process shown in FIG. 4 is performed by sequentially setting i to each integer value from 1 to MAX-P)ION (302).

45th! J represents the processing performed on the i-th phonetic symbol of one group. If the i-th phonetic symbol of the phonetic transcription of each word included in the group is all the same, then
The i-th phonetic symbol remains as it is, but if it is not, a marker "*" is inserted immediately after it.

As a result of the above processing, for example, the input dictionary entry in FIG.
As seen in the output dictionary entry of FIG. 5, a marker "*°" is inserted after a phonetic symbol for which a similar phonetic symbol exists.

(Effects of the Invention) As described above, according to the present invention, information indicating the position of similar phonetic symbols is added to dictionary items with similar pronunciations that are easily confused with each other, and based on this information, In the speech synthesis section, it becomes possible to synthesize speech while controlling the degree of emphasis of consonants and the like. Furthermore, the dictionary configured according to the present invention can be used in speech recognition to identify word candidates with similar pronunciations that are likely to be confused with each other.

[Brief explanation of the drawing]

FIG. 1 illustrates the words “
FIG. 4 is a diagram showing input dictionary entries for "Hospitality" and "Cheers." FIGS. 2 to 4 are flowcharts showing the flow of processing for adding information indicating the position of similar phonetic symbols of dictionary items with similar pronunciation to the input dictionary item. FIG. 5 shows the entries of a dictionary that is an embodiment of the present invention, and the words "hospitality" and "cheers" are added with information indicating the positions of similar phonetic symbols through the processing of FIGS. 2 to 4. FIG. 3 is a diagram of output dictionary entries for . 101... Input dictionary entry for the word "hospitality", 10
2... Input dictionary entry for the word "cheers", 501.
...Output dictionary entry for the word "hospitality", 502...
Output dictionary entry for the word "cheers".

Claims (1)

[Claims] In a word reading information storage dictionary consisting of dictionary items each having a phonetic notation consisting of at least a series of phonetic symbols, the phonetic notation of a certain dictionary item has a higher number of phonetic notations in the phonetic symbol series than the phonetic notation of another dictionary item. When corresponding positions have the same phonetic symbol and several corresponding positions have different but similar phonetic symbols, the certain dictionary entry contains information about the position in the phonetic transcription of the different but similar phonetic symbols. A dictionary for storing word reading information, characterized in that: is added.